Tbx3 encodes a transcriptional repressor that is important for diverse patterning events during development, and Tbx3 mutation in humans causes the ulnar-mammary syndrome. Here, we describe the identification of Tbx3 in array-based search for genes downstream Wnt/β-catenin that are implicated in liver tumorigenesis. Overexpression of Tbx3 is closely associated with the mutational status of β-catenin in murine liver tumors induced by Myc as well as in human hepatocellular carcinomas and hepatoblastomas. Moreover, Tbx3 transcription is activated by ectopic expression of β-catenin in mouse liver and in human tumor cell lines. Evidence that Tbx3 transcription is directly regulated by β-catenin is provided by chromatin immunoprecipitation and reporter assays. Although HepG2 cells stably transfected with Tbx3 display moderately enhanced growth rate, the dominant negative mutant Tbx3-Y149S drastically inhibits hepatoma cell growth in vitro and in vivo. Moreover, small interfering RNAs (siRNA) directed against Tbx3 inhibit anchorage-independent growth of liver and colon carcinoma cells. We further show that inhibition of Tbx3 expression by specific siRNAs blocks β-catenin–mediated cell survival and renders cells sensitive to doxorubicin-induced apoptosis. Conversely, ectopic expression of Tbx3 inhibits apoptosis induced by β-catenin depletion. Marked overexpression of Tbx3 in a subset of hepatoblastomas is associated with chemotherapy-resistant phenotype and unfavorable patient outcome. These results reveal an unsuspected role of Tbx3 as a mediator of β-catenin activities on cell proliferation and survival and as an important player in liver tumorigenesis. [Cancer Res 2007;67(3):901–10]
The Wnt/Wingless pathway plays a pivotal role in embryonic development and adult homeostasis by regulating cell fate, proliferation, and differentiation, and it is aberrantly reactivated in various human cancers (1). Mutations in adenomatous polyposis coli, Axin, and β-catenin genes that mimic Wnt activation result in the stabilization of β-catenin, which then moves to the nucleus and interacts with DNA-binding factors of the T-cell factor/lymphoid enhancer factor (Tcf/Lef) family to promote transcription of Wnt target genes. Depending on cell context, the selective expression of distinct sets of Wnt-responsive genes is strictly controlled by the interplay of signaling pathways. Increasing numbers of candidate β-catenin targets include cell cycle and growth regulators and proteins involved in cell-matrix interactions, migration and invasion, and in the regulation of the Wnt pathway.7
Recent studies have outlined the importance of Wnt/β-catenin signaling in regulating liver cell proliferation during development (2–4) and in governing essential functions in the adult liver (5–7). Moreover, aberrant reactivation of Wnt signaling is a predominant mechanism implicated in liver tumorigenesis. Mutations of the β-catenin and Axin genes leading to constitutive activation of β-catenin have been identified in hepatocellular carcinoma (HCC) and hepatoblastoma, and frequent overexpression of the Wnt receptor Frizzled-7 is a major early event in hepatocarcinogenesis (refs. 8–11; for reviews, see refs. 12, 13). Animal models have been instrumental in allowing the identification of β-catenin target genes in nondiseased or tumorous liver, including liver-specific enzymes involved in glutamine and nitrogen metabolism, such as glutamine synthetase (GS), ornithine aminotransferase, the glutamate transporter (GLT-1), and the cytochrome P450 enzymes CYP1A2 and CYP2E1 (5, 6, 14–17). However, c-myc has not been found thus far to be regulated by β-catenin in the liver context, and the oncogenic program triggered by Wnt signaling in hepatocarcinogenesis remains largely unknown.
We have shown previously that transgenic mice carrying a Myc oncogene controlled by woodchuck hepatitis virus (WHV) regulatory sequences were highly predisposed to liver cancer (18, 19). Activating mutations of β-catenin were found at significant rates in these HCCs, as for other Myc transgenic strains, suggesting that the survival functions of β-catenin might rescue tumor cells from Myc-induced apoptosis (8, 19, 20). Consistent with this notion, inactivation of p53 was found to represent an alternative oncogenic mechanism (9, 19). To explore downstream genetic programs activated by Wnt/β-catenin signaling in liver cancer, we employed microarray profiling and found that the T-box protein 3 (Tbx3) was specifically activated in murine tumors carrying mutant β-catenin.
The closely related genes TBX2 and TBX3 are members of the T-box gene family that play an important role in patterning events during development. Tbx3 is notably implicated in heart, limb, and posterior digit specification and in mammary gland development (21, 22). In humans, Tbx3 mutations are responsible for the ulnar-mammary syndrome (UMS), an autosomal dominant disorder characterized by upper limb deficiencies and apocrine/mammary gland hypoplasia (23). Tbx2/Tbx3 contain a conserved transcription-repression domain and can repress basal and activated transcription (24). Identification of p14ARF as a direct target of Tbx2/Tbx3 has linked these transcriptional repressors to the control of cell cycle and senescence and to the antiproliferative response delivered by the ARF-Mdm2-p53 pathway in tumorigenesis (25, 26).
In this study, we showed that the expression of Tbx3, but not Tbx2, was induced by mutant β-catenin in different systems, including normal murine liver, murine and human liver tumors, and human carcinoma cell lines. We cloned the TBX3 promoter and used a combination of chromatin immunoprecipitation (ChIP) and reporter assays that led to identify Tbx3 as a direct transcriptional target of the β-catenin/Tcf complex. We also showed that Tbx3 plays a crucial role in cell proliferation and survival. Inhibition of Tbx3 expression by RNA interference dramatically reduced anchorage-independent growth and abolished β-catenin protection over doxorubicin-induced apoptosis in tumor cells. Finally, we propose that strong overexpression of Tbx3 contributes to the chemotherapy-resistant phenotype of a subset of hepatoblastomas.
Materials and Methods
Transgenic mice and tumor samples. WHV/Myc transgenic mouse strains carry a woodchuck c-Myc or N-Myc2 transgene targeted to the liver by hepadnaviral regulatory elements and develop liver tumors almost invariably, as described previously (18, 19). WHV/Myc transgenic mice were genotyped by PCR on tail DNA and hybridization with WHV DNA as described (19). Animals were housed at the pathogen-free facility of Institut Pasteur and maintained according to the recommendations established by the French government (Services Vétérinaires de la Santé et de la Production Animale, Ministère de l'Agriculture). Mice carrying liver tumors were euthanized, and tumor samples were dissected free of liver parenchyma. Tumors and nontumoral livers were snap-frozen in liquid nitrogen and kept at −80°C.
RNA preparation and Northern blot analysis. Total RNA was extracted using the RNeasy system (Qiagen, Valencia, CA) and DNase treated (DNA-free Kit, Ambion, Austin, TX). Screening of β-catenin mutations was carried out as described previously (8). For Northern blot analysis, RNA samples (30 μg) were resolved by agarose gel electrophoresis and blotted onto Hybond N+ membranes. Blots were hybridized with Tbx3, Tbx2, and GS probes and with an 18S rDNA probe for loading control. Signals were quantified using the Storm 840 PhosphorImager (Molecular Dynamics, Sunnyvale, CA).
Murine cDNA arrays. Total RNA from 12 murine HCCs and matched nontumor livers was labeled using the Atlas Pure Total RNA Labeling System (Clontech, Mountain View, CA) according to manufacturer's recommendations. cDNA arrays spotted with the products of 1,185 genes (Clontech, ATLAS Mouse 1.2 Array) were used for differential hybridization analysis between tumors and matched livers. cDNA probe hybridization was done according to the manufacturer's recommendations. The array results were scanned using the Storm 840 PhosphorImager and analyzed using Atlas Image 1.01 software (Clontech Laboratories). Distance measurements and hierarchical clustering computations were done using Cluster and TreeView software. Differential gene expression between HCC and corresponding nontumoral liver was considered significant when signal ratio was >2.
Human tumor samples. Human HCC and hepatoblastoma specimens were obtained from patients undergoing surgical resection. Snap-frozen tumor tissues were collected from different French, Italian, and Chinese medical centers between 1998 and 2005. Most hepatoblastoma patients were enrolled in clinical trials of the International Society of Pediatric Oncology Liver Tumor Study Group. Informed consents were obtained at each hospital, and biological studies were approved by the review board of Institut Pasteur.
Total RNA was extracted from each sample using the RNeasy Kit (Qiagen) and treated with RNase-free DNase (Ambion).
Oligonucleotide microarray and statistical analysis. Fifty-five human HCCs, 24 hepatoblastomas, and 9 pools of nontumorous livers (5 from adults and 4 from children) were selected on the basis of high RNA quality as determined with the Agilent 2100 bioanalyzer (Agilent Technologies, Palo Alto, CA). Raw feature data from Affymetrix HG-U133A GeneChip microarrays were obtained for these samples. All samples were processed using the one-cycle target labeling protocol.827), which yielded log2 intensity expression summary values for each of the 22,283 probe sets. Each set (HCC and hepatoblastoma) was normalized independently. Additional statistical analyses were done with SPSS 11.0 package (SPSS, Inc., Chicago, IL). Comparison between groups was done using the Fisher's exact test. Probability of overall survival was determined using the Kaplan-Meier analysis and the log-rank and Breslow tests. Follow-up was closed at the time of death or last visit.
Plasmids. The reporter plasmid L-Tbx3-Luc carrying the TBX3 promoter was generated by PCR amplification of human genomic DNA sequences (University of California–Santa Cruz Genome Bioinformatics Site) extending 2,226 bp upstream and 619 bp downstream of the transcriptional start site of the human TBX3 gene (GenBank accession number NM_016569). We used the following primers: L-Tbx3F, 5′-CTAGCTAGCGAAACCCTGCAGTGACTTCCG-3′; and L-Tbx3R, 5′-GGAAGATCTGCTCGAAATAGACACTCCAGC-3′. The PCR product was cloned into pGL3-Basic (Promega, Madison, WI). A shorter promoter construct (S-Tbx3-Luc) extending 500 bp upstream of the transcription start site was generated by using the primers S-Tbx3F, 5′-CTAGCTAGCGCGAGCGGAGTGCAAGAGAGG-3′ and S-Tbx3R, 5′-GGAAGATCTCGGCGGCTCTAGAAGGTCG-3′. A third reporter construct carrying sequences upstream of the proximal TBX3 promoter (−2226 to −206) was derived from L-Tbx3 construct with the primers L-Tbx3F and Δ-Tbx3R: 5′-CCAGATCTCCTGTGAATATGTCA-3′. Mutation known to abolish Tcf binding (GTCAAAG → GCCAAAG) was introduced in the putative Tcf/Lef binding site using the Quick Change XL site-directed mutagenesis kit (Stratagene, La Jolla, CA). All plasmid inserts were verified by sequencing. The reporter plasmid containing the cyclin D1 promoter (-163CD1-Luc) was kindly provided by R. Pestell (Thomas Jefferson University, Philadelphia, PA). The expression vector for dominant stable β-catenin T41A was described previously (28). Full-length human Tbx3 cDNA was a generous gift of M. Van Lohuizen (The Netherlands Cancer Institute, Amsterdam, The Netherlands), and expression vectors encoding wild-type (wt) HA-Tbx3 and HA-Tbx3 Y149S mutant were provided by T.R. Brummelkamp (The Netherlands Cancer Institute, Amsterdam, The Netherlands; ref. 26). Tcf4 and dominant negative ΔNTcf4 expression vectors were kindly provided by H. Clevers (Netherlands Institute for Developmental Biology, Utrecht, The Netherlands).
Cell lines, transfections, and reporter assays. The human cell lines HepG2, Huh6 and Huh7 (hepatomas), 293 (kidney), SW480 and HCT116 (colon carcinomas), and U2OS (osteosarcoma) were maintained in DMEM with 10% fetal bovine serum. For reporter assays, semiconfluent cells in six-well plates were transfected with expression vectors for β-catenin, Tcf4, and/or ΔNTcf4 and 0.5 μg of luciferase reporter construct, using either calcium phosphate precipitation (293, SW480), Exgen (Euromedex, Souffelweyersheim, France; HepG2) or LipofectAMINE (Invitrogen, Carlsbad, CA; Huh6, HCT116). Luciferase activity was determined 48 h later. All experiments were repeated at least thrice. A thymidine kinase-β-galactosidase plasmid was cotransfected to normalize luciferase activity for transfection efficiency, and total DNA amount was kept constant by adding pcDNA3.
siRNA transfection. Cells were transfected 24 h after plating with small interfering RNAs (siRNA) against Tbx3 or β-catenin, or nonsilencing control (luciferase). siRNAs designed by selecting specific sequences of these genes were synthesized by Eurogentec (Seraing, Belgium). The target sequences were as follows: for β-catenin, 5′-AGC UGA UAU UGA UGG ACA G-3′; for Tbx3 (E), 5′-AUG GAG AUG UUC UGG GCU G-3′ and Tbx3 (F), 5′-GAG GAU GUA CAU UCA CCC G-3′; for Luc, 5′-CGU ACG CGG AAU ACU UCG A-3′. Cells were transfected with 100 nmol of siRNA per well using OligofectAMINE (Invitrogen).
Western blotting. Frozen mouse tissues or cells were lysed in chilled lysis buffer [50 mmol/L Tris-HCl (pH 7.5); 250 mmol/L NaCl; 0.5% Nonidet P-40; 250 mmol/L EDTA; 1 mmol/L DTT] supplemented with protease inhibitors (Roche, Basel, Switzerland). Whole extracts were resolved in 8% or 10% polyacrylamide gels and transferred to Hybond-C extra (Amersham Pharmacia Biotech, Buckinghamshire, United Kingdom). The following antibodies were used for immunodetection: mouse anti-β-catenin and mouse anti-glutamine synthase (BD-Transduction Laboratories, San Jose, CA), mouse anti-poly(ADP-ribose) polymerase (PharMingen, San Diego, CA), mouse anti-actin (Abcam, Cambridge, United Kingdom), and goat anti-Tbx3 (Santa Cruz Biotechnology, Santa Cruz, CA). Mouse and goat anti-immunoglobulin G conjugated to alkaline phosphatase were from Tropix (Applied Biosystems, Foster City, CA) and Sigma (St. Louis, MO). Immunoreactive proteins were visualized using the Western CDP-star kit (Tropix).
Chromatin immunoprecipitation assay. ChIP assay was conducted as described previously (29), with sheared DNA fragment size between 200 and 1,000 bp. Briefly, Huh6 cells were fixed with 1.1% formaldehyde, and nuclear extracts were isolated. The sonicated nuclear lysates were immunoprecipitated with a Tcf4 antibody (Upstate Biotech, Charlottesville, VA) or a β-catenin antibody (Santa Cruz Biotechnology). After purification of immunoprecipitated DNA, a 200-bp region of the Tbx3 promoter encompassing the putative Tcf site was amplified by PCR using the following primers: f55, 5′-AGCTCTATCCCCCAGCACTCG-3′, and r59, 5′-GAGAAAGCGAGAGCTCCTCGC-3′.
For control, two irrelevant regions of the Tbx3 locus located 10 kbp upstream or 10 kbp downstream were amplified using the primers: Upstream-F, 5′-AGCAGACTTCTGTAAAACAGG-3′ and Upstream-R, 5′-ACTCAGTAAGCTCTCTAAACG-3′; Downstream-F, 5′-GGGGGCCAGATCAGGCATGC-3′ and Downstream-R, 5′-GCTGCTGCTGGCAGCCTTTCC-3′.
For siRNA-coupled ChIP assays, Huh6 cells were transfected with siRNA for either β-catenin or luciferase. Cells were harvested 48 h later and processed as described above.
Cell proliferation and anchorage-independent growth assays. HepG2 cells stably expressing wt Tbx3 or the mutant Tbx3 Y149S were established by transfection using ExGen 500, followed by Geneticin selection. For cell proliferation studies, cells were harvested daily for 5 days and stained with trypan blue, and viable cells were counted on a hemocytometer. For in vivo tumorigenesis assays, 107 cells were implanted s.c. into the flanks of 6-week-old female athymic nu/nu mice (Charles River Laboratories, Wilmington, MA). Tumor size was determined 5 weeks later in groups of five to six mice by measuring tumor diameter.
For soft-agar assays, Huh6 cells were transfected with siRNA for β-catenin, Tbx3, or luciferase. After 24 h, cells were suspended in 0.3% agarose in DMEM supplemented with 10% fetal bovine serum. They were plated in triplicates in six-well dishes onto solidified 0.6% agarose-containing bottom layer medium at a density of 5 × 104 cells per well. Cultures were fed twice a week, and colonies were counted and photographed 10 days postplating.
TUNEL assay. U2OS cells were plated on glass coverslips, and cells at 60% confluence were transfected with β-catenin or empty vector together with siRNAs for Tbx3, β-catenin, or irrelevant control using LipofectAMINE. At 24 h after transfection, cells were treated with 500 ng/mL doxorubicin (Sigma-Aldrich, Lyon, France) and incubated for an additional 24 h. Cells were fixed with 4% formaldehyde in PBS for 15 min at room temperature and permeabilized with 0.1% Triton X-100 in PBS for 2 min on ice. Terminal nucleotidyl transferase–mediated nick end labeling (TUNEL) staining was done using the In Situ Apoptosis kit (Roche) according to manufacturer's instructions.
Tbx3 expression in murine HCC. We reported previously that WHV/Myc transgenic mice develop HCCs that carry frequent activating mutation of the β-catenin gene (8, 19). To identify β-catenin–induced genes that are involved in liver tumorigenesis, we generated gene expression profiles in 12 HCCs, including 6 cases harboring point mutation or interstitial deletion in the NH2 terminus of β-catenin and 6 cases with wt β- catenin. RNA from tumors and matched nontumoral livers was reverse transcribed, labeled, and hybridized to 1,185 gene cDNA arrays. The expression profile was compared between each tumor sample and corresponding liver, and data were expressed as T/NT ratio. We found 10 genes specifically deregulated (>2-fold) in the β-catenin mutant tumor category (Supplementary Table S1). Besides known β-catenin targets such as the glutamate transporter GLT-1 and the dynein light-chain genes, we found that Tbx3 was specifically up-regulated in HCCs bearing mutant β-catenin. Enhanced expression of Tbx3 was confirmed by Northern and Western blotting (Fig. 1). The mRNA levels of Tbx3 and the longer isoform Tbx3 + 2a (30) were increased 2- to 10-fold in tumors with mutant β-catenin compared with tumors with wt β-catenin and normal liver (Fig. 1A). The protein levels of Tbx3 showed parallel, significant variations between tumor samples (Fig. 1B). To validate these data, we analyzed Tbx3 expression in an additional set of 13 randomly selected HCCs from WHV/Myc mice. Compared with normal liver, Tbx3 mRNA levels were enhanced only in tumors carrying mutant β-catenin (Fig. 1C).
We then analyzed the expression of GS as a marker of activated Wnt/β-catenin signaling in the hepatic context (5). The Tbx3 and GS genes displayed strikingly similar expression patterns in WHV/Myc HCCs, both at transcriptional and protein levels (Fig. 1). Thus, the up-regulation of Tbx3 was associated to Wnt/β-catenin signaling. By contrast, expression of the related gene TBX2 remained constant in tumors and normal livers (Fig. 1C).
Tbx3 expression is tightly linked to activated Wnt/β-catenin signal in human liver tumors. These findings prompted us to investigate Tbx3 expression in human primary liver tumors. Oligonucleotide array (Affymetrix HG U133A) analysis was done in 55 HCCs and 24 hepatoblastomas specimens. Screening for β-catenin mutations revealed point mutations or interstitial deletions in the NH2-terminal region of β-catenin in 10:55 HCCs and 19:24 hepatoblastomas (Supplementary Tables S2 and S3). RNA from these tumors and control pools of nontumoral livers from adults and children were used for microarray analysis. Detailed data of global gene expression profiles will be published elsewhere for human HCC9
P. Pineau, A. Dejean, unpublished data.
C. Armengol, M.A. Buendia, unpublished data.
In human HCCs, expression of Tbx3 in the arrays was 2- to 10-fold higher in most tumors carrying mutant β-catenin compared with normal livers (Fig. 1D). Tbx3 was expressed at normal levels in a majority of tumors with wt β-catenin, but six HCCs in this group had highly elevated levels of Tbx3 mRNA. In these six cases like in the mutant β-catenin group, other known targets of β-catenin were overexpressed, including GS (Supplementary Fig. S1), pancreatitis-associated protein, regenerating islet–derived-1a, neuronal cell adhesion molecule, bone morphogenic protein (BMP) 4, and CYP2E1 (15, 17). It suggests that the Wnt/β-catenin pathway was activated through other means, such as overexpression of Frizzled-7 or loss-of-function mutations of Axin (9). Overall, in the 55 human HCCs analyzed, Tbx3 expression was significantly associated with deregulated β-catenin signaling (P = 0.002).
Similar analysis of 24 hepatoblastomas revealed the up-regulation of Tbx3 in 21:24 (87%) cases, including all tumors with mutant β-catenin and two tumors with wt β-catenin. We observed a 2- to 8-fold increase in the mutant β-catenin group compared with tumors carrying wt β-catenin (P = 0.007; Fig. 1D). Hepatoblastomas carrying mutant β-catenin displayed increased expression of known β-catenin targets, including GS (Supplementary Fig. S1), pancreatitis-associated protein, the Dickkopf homologue DKK1, BMP, and activin membrane-bound inhibitor homologue (BAMBI), BMP4, and the tyrosine kinase receptor EPHB2.11
Collectively, these results show that Tbx3 expression is tightly linked to the status of Wnt/β-catenin signaling in human and murine liver tumors. By contrast, no such association was evidenced for other T-box genes, including Tbx1, Tbx2, Tbx4, Tbx5, Tbx10, Tbx19, and Tbx21 (data not shown).
Tbx3 expression is induced by mutant β-catenin in mouse liver and human cell lines. Given the strong correlation between Tbx3 expression and β-catenin signal activation in liver tumors, we extended our studies to different nontumoral conditions. It has been shown that injection of recombinant adenovirus expressing the mutant S37A β-catenin strongly activated several Wnt target genes including GS (5). Northern blot analysis of livers from mice infected with Adβ-catS37A, or AdLacZ or AdGFP as controls, indicated parallel up-regulation of Tbx3 and GS in the livers of mice 48 h after injection of Adβ-catS37A (Fig. 2A). We next analyzed liver expression of Tbx3 in EAB/9K/β-catenin transgenic mice, which overexpress the deletion mutant ΔN131 β-catenin in the liver, kidney, and intestine and develop hepatomegaly soon after birth (14). On Northern analysis, expression of Tbx3 and GS was high in the livers of two transgenic mice aged 3 to 4 weeks, but it was barely detectable in matched nontransgenic livers (Fig. 2A). Both in Adβ-catS37A infection and in EAB/9K/β-catenin transgenic conditions, Tbx3 mRNA levels were increased by 2- to 3-fold.
To further address the regulation of Tbx3 by β-catenin, we investigated the effect of an active β-catenin mutant on endogenous Tbx3 expression in human cell lines that display no β-catenin activation. U2OS cells were transfected with the pcDNA-β-catT41A vector or empty vector, and significant up-regulation of Tbx3 was seen at 48 h after transfection, both at mRNA and protein levels (Fig. 2B). Similar data were obtained in Huh7 hepatoma cells (Supplementary Fig. S2). Thus, β-catenin–driven induction of Tbx3 might be independent of the cellular context.
β-catenin physically occupies and activates the Tbx3 promoter through Tcf/Lef. To determine whether Tbx3 is a direct transcriptional target of β-catenin, we adressed the association of β-catenin with the Tbx3 promoter by ChIP assays. A presumptive Tbx3 transcription start site located 965 bp upstream of the Tbx3 coding region was deduced from the 5′ ends of human and murine cDNA clones (accession numbers NM_016569 and XM_132317), and it was confirmed using primer extension (data not shown). Analysis of upstream sequences revealed a potential Tcf-binding site (GTCAAAG) at position −69 to −74 (see Fig. 3B), which is conserved in the murine sequence. Chromatin obtained from Huh6 hepatoblastoma cells, which carry mutant β-catenin G34V, was immunoprecipitated with either a Tcf4 antibody or a β-catenin antibody. In both cases, we could amplify the 200-bp region containing the Tcf-binding element in the proximal Tbx3 promoter (Fig. 3A). As control of the specificity of ChIP conditions, neither Tcf4 nor β-catenin was detected at distal upstream and downstream regions of the Tbx3 gene, and the Tbx3 promoter could not be amplified in the absence of the added antibody. Furthermore, β-catenin was no longer detected at the Tbx3 promoter after siRNA-mediated silencing of β-catenin in Huh6 cells, whereas irrelevant siRNA (siLuc) did not abolish β-catenin detection (Fig. 3A).
Then, the ability of β-catenin to activate Tbx3 transcription was investigated by reporter assay. We used PCR and human genomic DNA to clone 2,845 bp of promoter sequences upstream of the luciferase gene in pGL3 (L-Tbx3-LUC, Fig. 3B). We derived a shorter promoter construct covering 500 bp of the proximal promoter (S-Tbx3-LUC). Nucleotide substitution known to abolish Tcf binding (GCCAAAG) was introduced into the Tcf site (Lmut- and Smut-Tbx3-Luc). We also included in the analysis the ΔTbx3-Luc construct that is deleted of Tbx3 proximal promoter sequences (position −2,226 to −206, Fig. 3B). β-Catenin activity on Tbx3 transcription was examined in HCT116 and Huh6 cells, which carry activating β-catenin mutations. Cotransfection of the dominant negative mutant ΔNTcf4 with L- or S-Tbx3-Luc plasmids in HCT116 cells strongly decreased promoter activity in a dose-dependent manner, similar to the effects on the cyclin D1 promoter (Fig. 3C). By contrast, activity of the ΔTbx3-LUC reporter was low and insensitive to ΔNTcf4. Mutation of the putative Tcf-binding motif reduced L- and S-Tbx3 promoter activity by 2-fold and abolished sensitivity to ΔNTcf4. Similar data using the S-Tbx3 promoter were obtained in Huh6 cells (Fig. 3D). Moreover, in 293 cells, in which the Wnt pathway remains intact, the Tbx3 promoter was activated 3-fold by cotransfection of Tcf4 and β-catenin, but not by ΔNTcf4 and β-catenin (Supplementary Fig. S3). Taken together, these data provide evidence that Tbx3 is a direct transcriptional target of the β-catenin/Tcf complex.
Tbx3 can regulate proliferation and anchorage-independent growth of hepatoma cells. β-Catenin has been shown to increase cell proliferation and protect epithelial cells from suspension-induced apoptosis (anoikis; refs. 31, 32). We therefore tested whether Tbx3 might be implicated in the regulation of cell proliferation by β-catenin. To this end, we established HepG2 cell clones stably expressing either wt Tbx3 or the dominant negative Tbx3 Y149S mutant isolated from a UMS patient (26). Analysis of cell proliferation in representative HepG2 clones with similar transgene expression levels showed that expression of wt Tbx3 was associated with slightly enhanced growth rate, whereas mutant Tbx3 inhibited drastically cell proliferation (Fig. 4A). The ability of HepG2 cells expressing the Tbx3 mutant to form tumors in vivo was investigated by injecting nude mice with 10 × 106 cells. Groups of five or six mice were injected with either HepG2-Tbx3 Y149S or control HepG2 cells and examined 5 weeks later. In two independent assays, we found that a total of 10:12 mice injected with control HepG2 cells formed tumors, but only 1 of 10 mice injected with HepG2-Tbx3 Y149S developed a tumor (Fig. 4B).
Then, we examined the effects of Tbx3 on anchorage-independent growth of Huh6 cells in soft-agar assays. Cells were transfected with siRNA for either β-catenin, Tbx3, or luciferase as control and plated in soft agar 24 h later. Both β-catenin and Tbx3 siRNAs dramatically reduced the growth of cell clones, whereas robust clonal cell growth was seen in untransfected cells and in cells transfected with Luc siRNA (Fig. 4C). Similar data were obtained in four independent experiments, and reduction of β-catenin and Tbx3 expression in cells transfected with the corresponding siRNAs was verified by Western blotting (Fig. 4D). Similar results were obtained in HCT116, SW480, and 293 cells. Thus, impaired Tbx3 expression has potent inhibitory effects on the growth of cancer cells.
Tbx3 is involved in β-catenin–induced protection against apoptosis. It has been reported that β-catenin protects cells against p53-dependent apoptosis, and that Tbx3 has antiapoptotic activity, probably through down-regulation of p19ARF and subsequently p53 (33). To determine whether Tbx3 could mediate resistance to apoptosis conferred by β-catenin, we first examined the influence of Tbx3 on apoptosis induced by doxorubicin, a DNA damage–inducing agent known to induce p53-dependent apoptosis. In U2OS cells, transfection of β-cateninT41A and irrelevant siRNA (siLuc) reduced doxorubicin-induced apoptosis from 29 ± 10% to 19.25 ± 6.7% (mean ± SD), as determined by the frequency of TUNEL-positive cells (Fig. 5A). As expected, suppression of β-catenin by siRNA in cells transfected with mutant β-catenin restored apoptosis close to levels in control siLuc-treated cells. Strikingly, inhibition of Tbx3 by two different specific siRNAs in cells transfected with β-catenin T41A strongly increased doxorubicin-induced apoptosis to 47.5 ± 0.7% and 49 ± 1.4% (mean ± SD). Similar data were obtained in two independent experiments done in triplicates. Illustrative examples of TUNEL-positive cells are shown in Fig. 5B. These data were confirmed by using the caspase-mediated cleavage of poly(ADP-ribose) polymerase and Western blotting to evaluate apoptosis (Supplementary Fig. S4). It indicates that Tbx3 expression might be required for β-catenin–mediated protection against apoptosis.
We further analyzed the antiapoptotic activity of Tbx3 in HCT116 cells that express wt p53 and mutant β-catenin. Figure 5C shows that these cells were more resistant than U2OS cells to doxorubicin-induced cell death, but they were equally sensitive to apoptosis induced by siRNA-mediated depletion of β-catenin. At 48 h after transfection of siRNA against β-catenin, the ratio of cleaved/uncleaved poly(ADP-ribose) polymerase was increased 2.5-fold, and 3-fold with additional doxorubicin treatment. Interestingly, cotransfection of the Tbx3 expression vector with β-catenin siRNA decreased the level of apoptotic cells by 1.5-fold, both with and without doxorubicin treatment. Efficient down-regulation of β-catenin and Tbx3 by β-catenin siRNA and overexpression of Tbx3 in cells transfected with the Tbx3 vector were verified by Western blotting (Fig. 5C). Thus, expression of Tbx3 is required for suppression of apoptotic death by β-catenin in human tumor cells.
Clinical relevance of Tbx3 overexpression in hepatoblastoma. Given the antiapoptotic activities of Tbx3, we sought to investigate the relevance of Tbx3 in human liver tumors. We selected hepatoblastoma as a model because this tumor, unlike HCC, is currently treated by chemotherapy. The relationship between Tbx3 expression and various clinical features was investigated in the above-described panel of 24 hepatoblastomas (Supplementary Table S3). High-level Tbx3 expression (greater than or equal to the median Tbx3 expression value in hepatoblastomas, corresponding to a 4-fold increase compared with the median value in normal livers) was significantly associated with advanced tumor stage, poor response to chemotherapy, and shorter overall survival of patients (P < 0.05, Fisher's exact test; Table 1). The Kaplan-Meier survival estimate for patients in the high-level Tbx3 group was about 2-fold lower than for other hepatoblastoma patients (log-rank test, P = 0.021, Breslow test, P = 0.036; Fig. 5D). Thus, overexpression of Tbx3 might participate in tumor resistance to conventional chemotherapy, thereby influencing the outcome of hepatoblastoma disease.
|Variables .||Tbx3 < 4×|
|Tbx3 ≥ 4×|
|.||n = 16 .||n = 8 .||.|
|I or II||9||0||0.014|
|III or IV||7||7|
|No, solitary tumor||12||3||0.078|
|Metastasis at diagnosis|
|Response to chemotherapy|
|Variables .||Tbx3 < 4×|
|Tbx3 ≥ 4×|
|.||n = 16 .||n = 8 .||.|
|I or II||9||0||0.014|
|III or IV||7||7|
|No, solitary tumor||12||3||0.078|
|Metastasis at diagnosis|
|Response to chemotherapy|
Fisher's exact test.
PRETEXT stage, as defined by SIOPEL, is a staging system based on imaging at diagnosis that defines four categories depending on the number of liver sectors that are free of tumor (50). For one patient, no data were available.
In this study, we show that the T-box gene Tbx3 is a novel target of β-catenin implicated in the proliferation and survival of cancer cells. We show that Tbx3 expression is activated by Wnt/β-catenin signaling in murine as well as in human liver tumors, confirming the value of mouse models to study molecular events that occur in HCC (34). Moreover, we provide evidence that Tbx3 is a direct transcriptional target of the β-catenin/Tcf complex by showing that β-catenin physically occupies and activates the Tbx3 promoter, and that a single Tcf/Lef binding motif in the proximal Tbx3 promoter is critical for β-catenin responsiveness. These results add Tbx3 to the list of Wnt-responsive T-box genes, including the founding member T (Brachyury) and the Tbx3 orthologs optomotor-blind (Omb) in Drosophila and LvTbx2/3 in sea urchin (35–37). In this respect, it is of interest that deregulated Wnt/β-catenin signaling did not activate the close relative Tbx2 or any other T-box gene in the liver tumors analyzed in this study.
During development, Tbx3 has been implicated in inductive interactions at many stages of embryogenesis by its highly specific expression pattern. In particular, Tbx3 is expressed in the mesenchyme at the limb bud margins and in the mammary buds, correlating with defects in anteroposterior patterning and mammary gland induction in Tbx3-deficient mice and in UMS patients (22, 23, 38). Strikingly, published evidence placing Tbx3 downstream of β-catenin in these processes is limited to the cooperative effect of Wnt and fibroblast growth factor signals in mammary gland initiation (39). Previous studies rather pointed to the role of Sonic hedgehog and BMP-mediated signals in regulating Tbx3 expression (21, 40–42). Notably, it has been shown that Tbx3 is a direct target of BMP Smads (43). Therefore, Tbx3 expression seems to be governed by a complex network of signals. The question of whether regulatory cross-talks between Sonic hedgehog, Bmp, and Wnt pathways also operate in oncogenic processes to regulate Tbx3 expression warrants further investigation.
In this study, activation of Tbx3 by the Wnt/β-catenin pathway was initially recognized in Myc-induced murine liver tumors that carry second-hit mutations in the β-catenin gene (8, 19). Wnt/β-catenin signaling is known to protect cells from Myc-induced apoptosis, thereby facilitating transformation (44). Strikingly, other studies have pointed to similar activity of Tbx3 in counteracting Myc-induced apoptosis (33). Our finding that a Tbx3 dominant negative mutant could potently inhibit hepatoma cell growth in vitro and in vivo provides evidence that Tbx3 mediates, at least in part, the growth-promoting effects of β-catenin in cancer cells. Moreover, siRNA-mediated depletion of Tbx3 dramatically inhibits the growth of hepatoblastoma cells in soft agar. We also show that Tbx3 expression is mandatory for the protective effect of β-catenin over p53-dependent apoptosis induced by doxorubicin, and that the overexpression of Tbx3 rescues cells from apoptosis induced by β-catenin depletion. Thus, Tbx3 seems to be a critical determinant of cellular responses to proliferative and antiapoptotic signals delivered by β-catenin. These effects can be explained by transcriptional repression of the ARF tumor suppressor gene by Tbx3, leading in turn to the down-regulation of p53 and the activation of Myc-induced proliferation and transformation, but not Myc-induced apoptosis (26, 45).
To address the question of how these activities manifest in vivo, we used hepatoblastoma, a tumor that responds generally well to current chemotherapy regimens such as cisplatin and doxorubicin, although refractory cases (20–30% of patients) with advanced tumor stage or metastasis have poor survival rates (46). The presence of β-catenin mutations in a majority of hepatoblastomas is not associated with disease progression or any tumor characteristic (47). We show that high overexpression of Tbx3 in a subset of hepatoblastomas carrying mutant β-catenin is correlated with advanced tumor stage, poor response to chemotherapy, and unfavorable prognosis. Overexpression of Tbx3 is not restricted to liver cancer because it has been described in breast and ovarian cancer (30, 48) and in ovarian endometrioid adenocarcinomas characterized with respect to mutations affecting the Wnt/β-catenin pathway (49). Based on the key role of Tbx3 in carcinogenesis, inhibition of Tbx3 activity could be an effective therapeutic strategy for human cancer.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
Grant support: Ligue Nationale contre le Cancer (Carte d'Identité des Tumeurs Program), Association pour la Recherche sur le Cancer, Ecole Normale Supérieure de Lyon (C. Labalette), European Association for the Study of the Liver Sheila Sherlock Fellowship (C. Armengol), and GIS-Institut des Maladies Rares (S. Cairo).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank Laurence Lévy and Oliver Bishof for stimulating discussion and Daniela Geromin and Jean-Yves Coppée for help with RNA quality control and quantitative PCR. We thank the French, Italian, and Chinese medical centers that provided us with tumor specimens, in particular Drs. M. Fabre (Hôpital de Bicêtre, Le Kremlin-Bicêtre, France), B. Terris (Hôpital Cochin, Paris, France), B. Turlin (Centre Hospitalier Universitaire Pontchaillou, Rennes, France), V. Mazzaferro (Tumor National Institute, Milan, Italy), and L.X. Qin (Fudan University, Shangai, China). We are grateful to Bert Vogelstein, Thijn Brummelkamp, Maarten van Lohuizen, Hans Clevers, and Richard Pestell for providing reagents.